UPTEC W11 017

Examensarbete 30 hp
December 2011

The Overhead Line Sag Dependence

on Weather Parameters and Line
Current

Elisabeth Lindberg
Abstract
The overhead line sag dependence on weather parameters and line
current

Elisabeth Lindberg

As the demand for energy increases, as well as the demand for renewable energy,
Vattenfall, as network owner, receives many requests to connect new wind power to
the grid. The limiting factor for how much wind power that can be connected to the
grid is in this case the maximum current capacity of the overhead lines that is based on
a line temperature limit. The temperature limit is set to ensure a safety distance
between the lines and the ground.

This master thesis project is a part of a research project at Vattenfall Research and
Development that is examining the possibilities of increasing the allowed current on
overhead lines in order to be able to connect more wind power to the existing network.
Measured data from two overhead lines in southern Sweden is analyzed and the
internal relations between the measured parameters are examined. The measured
parameters are overhead line sag, line temperature, ambient temperature, solar
radiation, wind speed and line current.

The results indicate that there is a big load margin that could be utilized to increase the
maximum current as long as further work could show that low winds at line height
correlates with low wind at nacelle height. The results show that the sag versus line
temperature is approximately linear within the measured temperature range. This
means that a real-time-monitoring system measuring the line temperature should give
adequate knowledge of the line position to ensure the safety distance. A model for the
line temperature as a function of insolation, current, ambient temperature and wind
speed has been estimated for one of the lines. Simulations show that a sudden increase
in current at a worst-case scenario would give the operators about ten minutes to react
before the line reaches the temperature limit.

Keywords: Overhead lines, dynamic rating, real-time monitoring, conductor

temperature model, sag

Department of Information Technology, Uppsala University, Box 337, SE- 751 05

Uppsala
ISSN 1401-5765
Referat
Beroendet av vderparametrar och strm hos nedhng p luftledningar

Elisabeth Lindberg

D energibehovet i samhllet kar samt intresset fr frnyelsebar energi har Vattenfall

brjat f vldigt mnga frfrgningar om nya vindkraftanlggningar som vill ansluta
till Vattenfalls nt. Den begrnsande faktorn fr hur mycket vindkraft som kan
anslutas till ntet blir i detta fall den maximalt tilltna strmmen i luftledningarna,
vilken baseras p en maximal tillten temperatur p ledningen. Denna maxgrns r satt
s att den garanterar att skerhetsavstndet mellan ledningen och marken inte blir fr
litet.

Det hr examensarbetet r en del i ett forskningsprojekt p Vattenfall Research and

Development som undersker mjligheten att ka den maximalt tilltna strmmen i
luftledningar fr att kunna ansluta mer vindkraft till ntet. Mtdata frn tv
luftledningar i sdra Sverige har analyserats och sambanden mellan de mtta
parametrarna har studerats. De mtta parametrarna r linans nedhng, lintemperatur,
lufttemperatur, solinstrlning, vindhastighet och strm.

Resultaten tyder p att det finns en stor marginal som skulle kunna utnyttjas fr att ka
den maximalt tilltna strmmen i linan, under frutsttning att ven en korrelation
mellan lg vind p lin-hjd och lg vind p vindkraftverks-hjd kan pvisas i fortsatta
studier. Resultaten visar att linans nedhng har ett approximativt linjrt frhllande till
lintemperaturen. Detta betyder att ett realtidsvervakningssystem med
lintemperaturmtning skulle kunna bestmma linans position med tillrcklig
noggrannhet fr att kunna garantera att skerhetsavstndet bibehlles. En modell fr
lintemperaturen som funktion av strm, solinstrlning, lufttemperatur och
vindhastighet har anpassats fr en av linorna. Simuleringar har visat att vid en pltslig
strmkning vid ett vrsta tnkbara scenario skulle operatrna ha ungefr tio minuter
p sig att reagera innan linan nr temperaturgrnsen.

Nyckelord: Luftledningar, dynamisk rating, realtidsvervakning, lintemperaturmodell,

nedhng

Institutionen fr informationsteknologi, Uppsala universitet, Box 337, SE- 751 05

Uppsala
ISSN 1401-5765

ii
Preface
This master thesis project was performed at Vattenfall R&D in Rcksta, Stockholm,
from November 2010 to June 2011, within the Master of Science program in Aquatic
and Environmental Engineering at Uppsala University. Supervisors at Vattenfall were
Lovisa Stenberg and Urban Axelsson. Subject reviewer was Bengt Carlsson,
Department of information technology at Uppsala University and examiner was Allan
Rodhe, Department of earth sciences at Uppsala University. This master thesis has
been a part of the research project Wind cooled OH lines at Vattenfall.

First I would like to thank Vattenfall R&D for giving me this opportunity of working
with this interesting project. Many thanks to Urban Axelsson and Lovisa Stenberg for
their commitment and guidance during this project. Their support has been of great
importance for me and this project. Thanks to Bengt Carlsson for ideas and support
with the line temperature modelling. Thanks to Stefan Lindstrm for helping me with
the sag modelling. Thanks to Elisabeth Larsson, Martin Tillenius and Marcus Holm
for their help with the Matlab programming. Thanks to Gunnar Erixon for guiding at
the visit at the operation center in Trollhttan. Thanks to Tor Johansson and Arne
Bergstrm especially for help with preparing the presentation and thanks to Peter
Schelander for preparing weather data and for good advice.

Copyright Elisabeth Lindberg and the Department of Information Technology,

Uppsala University
UPTEC W 11 017, ISSN 1401-5765
Printed at the Department of Earth Sciences, Geotryckeriet, Uppsala University,
Uppsala, 2011

iii
Populrvetenskaplig sammanfattning
Beroendet av vderparametrar och strm hos nedhng p luftledningar

Elisabeth Lindberg

D energibehovet i samhllet kar samt intresset fr frnyelsebar energi har Vattenfall

brjat f vldigt mnga frfrgningar av nya vindkraftanlggningar som vill ansluta
till Vattenfalls nt. Lngs vissa ledningsstrckor p ntet rcker inte kapaciteten till fr
att vindkraftverken ska tilltas att ansluta direkt till befintligt nt. Om kapaciteten p
ledningarna verskrids blir ledningarna varmare n de r designade att klara av. D
blir nedhnget mellan ledningsstolparna stort och det skerhetsavstnd som mste
uppfyllas mellan ledningen och marken riskerar att bli fr litet.

En lsning p detta problem skulle kunna vara att bygga ut kapaciteten genom att
bygga nya ledningar. Problemet r att detta r en lng och dyr process som mste
genom mnga instanser innan ledningar kan godknnas och byggas. En annan lsning
r att se om kapaciteten i ledningarna egentligen r underskattade. Detta r grunden till
det forskningsprojekt p Vattenfall Research and Development som detta
examensarbete r en del av.

Kapaciteten i ledningarna begrnsas allts av ett minsta avstnd till mark som enligt
lag mste upprtthllas. Avstndet till marken minskar nr ledningen blir varm
eftersom att materialet i linan utvidgas och linan blir lngre. Temperaturen i linan
bestms av hur mycket strm som gr i ledningen och av det omgivande vdret.
Lufttemperatur, sol och vind har stor pverkan p linans temperatur. Eftersom att
vindkraftverken som ska anslutas inte stts igng fr energiproduktion under en viss
vindstyrka r tanken att belastningen p ledningarna inte kommer att kas s lnge
vinden r svag och att nr vinden r stark kommer den att kunna kyla av ledningen
tillrckligt mycket fr att jmna ut den uppvrmning som kommer av den kade
strmmen.

Fr att testa om detta r mjligt har man inom forskningsprojektet satt upp tv stycken
mtsystem p tv ledningar i sdra Sverige dr man mter vdret, strmmen,
ledningens temperatur samt avstndet mellan ledningen och marken. Syftet med
examensarbetet har varit att finna sambanden mellan de mtta parametrarna genom att
analysera mtdata och att ska i litteratur.

Examensarbetet har lett till att en frenklad modell fr ledningens temperatur som
funktion av strm och vder har tagits fram samt en modell fr storleken p
ledningens nedhng som funktion av ledningens temperatur p det spann dr
mtningarna har genomfrts. Med modellen har man kunnat simulera ledningens

iv
temperatur vid olika situationer. Med simuleringarna har man underskt hur lng tid
det skulle ta efter en kraftig strmkning i ledningarna under smsta tnkbara
vderfrhllanden innan ledningens temperatur ntt maxnivn. Detta motsvarar den tid
en operatr i driftcentralen skulle ha p sig att agera vid en kraftigt uppmtt
strmkning. Man har ven underskt hur stor pverkan de olika vderparametrarna
har relativt varandra p temperaturkningen i ledningen och hur olika tjocklekar p
linan pverkar reaktionstiden samt hur stor den verkliga outnyttjade kapaciteten har
varit utgende frn mtdata.

Resultaten tyder p att den outnyttjade kapaciteten i linan r stor (upp till dubbla
kapaciteten), men inte frrn analys av de fortsatta mtningarna har gjorts gr det att
veta hur stor marginal som finns under den varmaste och kallaste delen av ret. Det
terstr ven att underska om de tillfllen som ger lgst verklig kapacitet
sammanfaller med lg vindhastighet vid hgre hjd.

v
Table of Contents Page

1 INTRODUCTION 1

1.1 Purpose of the master thesis project 2

1.2 About the report 2

2 BACKGROUND 3

2.1 The Swedish grid 3

2.2 Overhead lines 4
2.3 ACSR conductors 4
2.4 The monitoring at operation centers 5

3 MEASUREMENTS 6

3.1 Two measurement sites 6

3.1.1 Site OL9 6
3.1.2 Site ZL8 6
3.2 Measurement system 6
3.3 Measurement data range 9

4 THEORY 11

4.1 Sag theory 11

4.1.1 Thermal elongation 13
4.1.2 Other line elongation causes 13
4.1.3 Coupled spans 14
4.2 Conductor temperature model 15
4.2.1 Current heating 17
4.2.2 Solar heating 18
4.2.3 Radiative cooling 18
4.2.4 Wind cooling 20
4.3 Thermal rating 22
4.4 Line uprating 23
4.4.1 Redefining weather conditions 23
4.4.2 Dynamic Rating and real-time monitoring 23
4.4.3 Probabilistic methods 24
4.5 Correlation between weather parameters 24
4.6 Grey-box modelling 24

vi
5 METHODS 26

5.1 Data analysis 26

5.1.1 The system 26
5.1.2 Sag-temperature relationship analysis 27
5.1.3 Modelling line temperature with a multiple linear regression model 28
5.1.4 Calibrating the dynamic model of line temperature in Matlab 28
5.1.5 Evaluation of dynamic line temperature model 30
5.1.6 Examination of the system behavior using the dynamic model 31
5.2 Estimating a dynamic line temperature model for OL9 31

6 RESULTS 32

6.1 Sag versus line temperature, ZL8 32

6.2 Result from multiple regression 34
6.3 Calibration results for dynamic line temperature model 35
6.4 The influence of weather parameters on conductor temperature change 40
6.5 Time constants and step responses 41
6.6 Worst-case simulation 44
6.7 Possible rating of the two lines 46

7 DISCUSSION 48

7.1 Analysis of ambient conditions and line temperature 48

7.2 The usefulness of the line temperature model 48
7.3 Possible weaknesses of the dynamic line temperature model 48
7.4 Implications from results of possible ratings 48
7.5 The simplified wind model term 49
7.6 Time constants and times for reaching temperature limit 49
7.7 Variations along the line 50
7.8 Sag-line temperature correlation 50
7.9 Suggestion of uprating method 51

8 REFERENCES 53

vii
Appendices Number of Pages

APPENDIX pp

1 - Measured line temperature plotted versus other measured parameters 5

2 Profile images of measured spans 2

viii
1 Introduction
As the demand for energy increases, as well as demand for renewable energy,
Vattenfall, as network owner, receives many requests to connect new wind power to
the grid. Since the wind power has to be located at places where the wind is
adequately strong the system has to adjust to the location of the wind farm rather than
the other way around. This means that in order to connect the wind farms to the grid
new overhead lines must sometimes be built or old ones reinforced. The building of
new lines is a very costly and time-consuming process as it can take years for
applications to go through the regulation system. Public opinion of new overhead lines
can also slow down the process. Reinforcing old lines is also very expensive as the
lines must be taken out of service for an extensive time. Therefore it is of big interest
for Vattenfall to examine if the capacity of the lines is underrated.

Overhead line ratings (maximum current capacity) for these lines are based on the
maximum allowable temperature of the conductor. The maximum allowable
temperature is a temperature that is set so that the conductors will not at any condition
during their lifetime get closer to the ground than the minimum electrical clearance
distance, which is set by regulations. The minimum clearance distance is set to ensure
public safety. A person getting too close to a high voltage electrical conductor could
result in a flashover that could lead deadly currents through the passing person. This is
why the safety aspect of overhead lines is so important and why an uprating of lines
should be done carefully, backed up by both measurements and theory.

The sag of the line increases when the line elongates due to increasing temperature,
mechanic load and time. The temperature of the conductor is dependent on conductor
properties, line current and ambient conditions.

Todays line ratings are based solely on worst case assumptions of weather conditions.
This means that the full capacity of the line is not used. This master thesis project is
one part of a research project at Vattenfall which intends to uprate overhead lines for
wind power insertion to the grid. If one can find a correlation between wind speed at
nacelle (the generator house of the wind turbine) height and wind speed at line height
wind power could be installed and connected to the grid to utilize a larger part of the
actual capacity, since high wind gives high load on the line at the same time as high
wind cools the line efficiently. The wind power could also be cut off to lower the load
on the line in emergency cases when wind speed at line height would not cool the line
enough as a means to ensure the line safety.

1
1.1 Purpose of the master thesis project

The purpose of this master thesis was to investigate the relationships between the
parameters that affect the line temperature and sag of overhead lines using measured
data and literature search and to increase the knowledge in general about the sag and
line temperature behavior at Vattenfall.

Specific aims were to:

Identify correlations between the measured parameters.
Highlight differences between the two overhead lines.
Find existing calculation procedures and methods in literature.
Give recommendations concerning the load capacity as a function of all involved
parameters.

1.2 About the report

This master thesis report will first go through some general theory about the Swedish
grid (to put the measurements into context), about overhead lines in general and
ACSR (Aluminum Conductor Steel Reinforced) conductors. The procedure at the
operation center regarding the use of todays ratings will be shortly described. The
measurement chapter deals with the data collecting measurement system. The theory
chapter is the outcome of the literature search and gives equations for sag behavior
and line temperature change behavior as well as a brief introduction to grey-box
modelling. Methods of uprating are also described in this chapter. The methods
chapter describes the procedure of estimating models for line temperature and sag.
The report finishes with results from model estimation and simulations and discussion
of results as well as some suggestions of future work.

2
2 Background
2.1 The Swedish grid
Transmission
network

Regional
network

Local
network

Figure 2.1: Voltage levels of the Swedish electrical grid, (modified from [12])

The Swedish electrical grid is divided into different voltage levels (Figure 2.1) [12].
The transmission network has the highest voltage levels, at 400kV and 220kV, and is
used for long-distance transportation of electricity. This part of the network is owned
and operated by Svenska Kraftnt, owned by the Swedish state.

The electricity is transformed down to lower voltage levels in transformation stations

and is then called the regional network. The regional networks are owned and operated
by the bigger energy utility companies in Sweden of which Vattenfall is one. The
voltage level for the regional network is 40kV to 130kV. The regional network
transports electricity from the transmission network to the local networks or directly to
large industrial electricity consumers.

3
The local network has voltage levels below 40kV. The local network distributes the
electricity to the final consumers such as households. Local networks in Sweden are
owned by several local operators.

2.2 Overhead lines

The most used method of electricity transport in transmission and sub-transmission

systems is with overhead lines. The alternative to overhead lines is underground
cables which are more expensive than overhead lines and therefore not as common.

Figure 2.2: Overhead lines with different structures, [8] and [4].

Overhead lines can be built with steel towers or wooden poles (Figure 2.2) and can
have different conductor materials. Higher voltage levels on the line require bigger
and stronger structures. The conductors are held up by the towers via insulator strings.

Overhead lines must always have a minimum electrical clearance due to public safety.
The safety distance for 130 kV overhead lines in Sweden is 7,4 m above ground for an
area with detailed planning and 6,4 m for areas without detailed planning [19].

2.3 ACSR conductors

There are several types of conductors that are used for overhead lines. The most
common is the ACSR (Aluminum Conductor Steel Reinforced) conductors [3]. Other
types of conductors are all aluminum, copper and aluminum alloy. The desirable
properties of a conductor are that it has low resistivity and has sufficient strength to be
able to carry its own weight and additional load for its entire lifetime. Copper has
lower resistivity than aluminum, which makes it a good electrical conductor, but has
greater density, which means that the tower constructions and the insulators have to be
heavier for copper conductors. This, together with the fact that aluminum is cheaper
than copper, makes aluminum the more popular choice for overhead line design.

4
Aluminum is however a weaker material than copper and this is why the core of steel
is added to the conductor, to increase its strength. [11]

The conductor is not solid but built up from aluminum and steel strands that are
spiraled. One reason for stranding is the need for flexibility in the line. [11].

Figure 2.3: Cross-sections of three types of ACSR conductors, [10]

There are many different ACSR-conductors. The conductors that were measured on
are of types 42/7 and 54/7. The first number stands for number of aluminum strands
and the second for number of steel strands. In Figure 2.3 cross-sections of some ACSR
conductors are shown.

The different build-up of aluminum and steel strands give the conductors different
properties. The aluminum-to-steel area ratio affects the elastic modulus, the thermal
elongation coefficient, the resistance of the conductor and the magnetic losses.

2.4 The monitoring at operation centers

The work of everyday-monitoring of overhead line operation is done at the various

operation centers. A visit at the operation center in Trollhttan gave this description of
how the operation of monitoring overhead lines is currently run.

Today the current capacity of a line (the rating) is tabulated for different ambient
temperatures and constant wind speed of 0.6 m/s for a maximum allowed line
temperature of 50C. A computer monitoring system that receives current data from
the stations along the line uses two standard values for thermal rating, one for summer
and one for winter. For 130kV lines the summer rating is based on an ambient
temperature of 30C and the winter rating on 10C. If the monitoring system spots a
current value in a station that is getting close to the rating an alarm goes off on the
operators screen. The operator then decides what to do about the emergency.
Overloads usually occur on cold winter days and then the operator knows from
experience that the line can handle somewhat higher currents than the tabulated
values.

5
3 Measurements
3.1 Two measurement sites

The measurement data used in this project comes from two 130kV overhead lines in
southern Sweden, one in Brlanda and one in Bjrnsen. The measurements began in
July 2010, before the start of the master thesis project, and are still ongoing. The
measured parameters are current, solar intensity, wind speed and direction, ambient
temperature, relative humidity, conductor temperature and distance to ground.

The two measurement sites were selected on the basis of several important factors.
The sites had to be located so that installation of equipment could go smoothly with
easy access by car and also access to power to be able to supply heating of the
measuring equipment. A good cellular network coverage was necessary for easy data
transfer. High loading on the lines was preferred since the reason for measuring is
trying to uprate the lines. If the loading already is so low that the full capacity is not
used, then uprating would be unnecessary. An important factor on top of this was that
the site had to have good shielding from wind to represent a hot spot on the line.

3.1.1 Site OL9

The line OL9 that runs through Brlanda was selected as previous simulations of the
load flow had shown that this line possibly would reach limiting line loadings when
connecting more wind power to the grid. A profile image of the chosen span for
measurements is shown in Appendix 2. OL9 has a diameter of 27.8 mm and the span
length at the site is 128 m.

3.1.2 Site ZL8

The line ZL8, which runs through Bjrnsen, was chosen as the second line. It was
chosen by means of its closeness to the coast, voltage level, dominating wind
direction, the possibilities to perform planned interruptions, line loading and
similarities and differences in line properties compared to OL9, such as diameter and
span lengths. A profile image of the span is included in Appendix 2. The diameter of
ZL8 is 39.2 mm and the span length is 247 m.

3.2 Measurement system

The measurement systems that have been used are a Cordina box and a weather
station. The measurement systems are identical for the two sites except for power
supply system.

6
Figure 3.1: The measurement system Cordina installed on the line and the
weather station on a pole beside the line

The Cordina box (Figure 3.1 and Figure 3.2) is a system which is installed on the line
itself and measures line temperature, ambient temperature, current and distance to
ground. The distance to ground is measured with a laser that reflects on the ground.
The measurement system logs the data and communicates it to a computer at
Vattenfall.

7
Figure 3.2: The inside of the Cordina measurement system

The weather station is located 6 meters beside the line (Figure 3.1 and Figure 3.3) and
was placed at the height the Cordina box on the line at the time of installation. The
wind measurement is done with a three-dimensional ultrasonic anemometer. The
insolation sensor is placed so that no object will shadow it during times of high
insolation. The ambient temperature and relative humidity sensors have radiation
protection with ventilation that should limit the impact of solar radiation. The ambient
temperature measurement from the weather station has in the master thesis project
only been used in comparison to the ambient temperature measured with the Cordina
box as to verify that the time tagging between the two systems has been correct. The
relative humidity data has not been analysed.

The aspiration was to have data from two different lines to compare and also to have
measurements from all seasons of the year. Unfortunately there were some technical
difficulties with the Cordina measurements and communication system and therefore a
lot of data is missing for July and August 2010 at Bjrnsen. Complete data sets have
not yet (May 2011) been obtained in Brlanda because the Cordina measurement has
not been working and the line could not be taken out of service for repairs until late
March 2011. The repairs still did not succeed in fixing the line temperature
measurements which means that no line temperature data for Brlanda have yet been
obtained.

After having shown unreasonal behaviour of freezing at one value the laser
measurements in Bjrnsen were improved with the placing of a white board on the
ground under the box. This was done some time in October 2010.

8
Figure 3.3: Close-up of the weather station

3.3 Measurement data range

The monthly max- and min-values for the measured parameter are displayed in Table
3.1. Measured data for periods not mentioned in Table 3.1 have also been collected
but not taken into account within this thesis and are therefore not displayed in the
table. High current was wanted as well as high line temperatures as the higher
temperatures are of most interest within this project.

Notation for Table 3.1:

I Current
V Perpendicular wind speed
S Insolation (Note that S also denotes span length in chapter 4.1)
Ta Ambient temperature
Tc Conductor temperature

9
Table 3.1: Range for collected measurement data
2
I (A) V (m/s) S (W/ m ) Ta (C) Tc ( C)
Min Max Min Max Max Min Max Min Max
Bjrnsen, ZL8 July 128 672 0.0 7.2 960 11 27 13 42
August 177 461 0.3 3.9 1012 12 20 13 32
September 12 721 0.0 5.1 931 -1 22 -1 34
October 110 913 0.0 5.1 651 -7 17 -3 40
April 6 889 0.0 6.0 1034 -2 22 0 38

Brlanda, OL9 April 3 223 0.0 3.9 714 1 21 - -

10
4 Theory
4.1 Sag theory

The shape of a hanging line attached in its two ends can be described as a function of
horizontal tension and weight per unit length [3].

Figure 4.1: Illustration of conductor sag, [3]

Figure 4.1 illustrates the meaning of sag, span length and line length, where

D = Sag (m)
S = Span length (m)
L = Line length (m)
H = Horizontal component of tension (N)
T = Total tension (N)
w = weight per unit length of conductor. (N/m)
x = horizontal distance from lowest point (m)
y(x) = vertical distance from lowest point at x (m)

The shape of the hanging line can be described with an approximate parabolic
equation or with a more exact hyperbolic catenary equation. The error due to parabolic
approximation is very small except for very long, steep or deep spans. The parabolic
equation, has the advantages that it easily shows the relationships between sag,
tension, weight and span length [3].

11
The catenary equation is given by:

H w x (4.1)
y ( x) = cosh 1
w H
and the parabolic equation reads:

w x2 (4.2)
y ( x)
2 H

As the overhead lines are placed in the terrain the attachment points are not always
placed on the same height. A span that has both attachment points at the same vertical
level is called a level span. A span with its attachment points on different heights is
called an inclined span. The calculation of sag for inclined spans is more complex than
for level spans as the lowest point of the curve moves with changes in conductor
length. Increasing length of the conductor moves the lowest point closer to the
midpoint between the poles.

Figure 4.2: Illustration of conductor sag for inclined span [3]

The horizontal position of the lowest point for an inclined span can be described by:

S H h2
x r = sinh 1 (4.3)
2 w H sinh S 2
w H w
where h is the difference in height between the two attachment points.

Figure 4.2 shows an inclined span. The sag for an inclined span will not describe the
lowest point of the conductor but is defined in the middle of the span.

12
For a level span, using the parabolic approximation, the sag can be described by:

w S 2 (4.4)
D=
8 H

It is apparent from (4.4) that sag is proportional to the square of the span length.

Sag can also be approximated as a function of line and span length as:

3 S (L S ) (4.5)
D=
8

These equations assumes full flexibility of the conductor [3].

4.1.1 Thermal elongation

The increase of sag with conductor temperature is due to expansion of the conductor
material. The elongation can be described with a linear relationship [3]

L
= AS Tc
L (4.6)

where AS is the coefficient of linear thermal elongation.

The coefficient of linear thermal expansion is specific for each line as it depends on
the aluminum-to-steel area ratio. Aluminum elongates with twice the rate as steel so
the more aluminum the larger the coefficient of thermal expansion. These values are
tabulated for different conductors [13].

When the conductor elongates due to temperature change the tension within the line
also changes. In order to calculate the result accurately one needs to solve both
equations for length and tension simultaneously [3].

Thermal elongation is described as an elastic elongation, which means that the process
is reversible.

4.1.2 Other line elongation causes

The conductor is also subject to other elongation causes than elastic thermal
elongation such as long-time creep and mechanical ice and wind loads. High loads as
well as creep causes plastic elongation (irreversible elongation) and changes the
elastic modulus of the conductor.

13
When overhead lines are designed and installed they are installed so that the minimum
clearance is to be ensured under all conditions that can apply under the lifetime of the
conductor. The clearance must therefore hold for all the predictable ice and wind
loads, for the maximum allowed conductor temperature and for the long-time creep
(Figure 4.3).

Figure 4.3: Illustration of different causes to sag during a lines lifetime, [3]

Values for final elastic modulus after wind and ice loading can be found in tables
specified for each line type, see [13].

4.1.3 Coupled spans

The overhead line conductors are attached to the towers via insulator strings. Usually
the insulator strings are somewhat free to move. The insulator moves when it is
subjected to unequal forces at the bottom point, where the conductors are attached.
The unequal forces arise when there is a difference in tension between spans. This
means that any difference in tension between adjacent spans is equalized. [3] This also
means that the sagging of one span is not isolated from the other spans and cannot
accurately be calculated separately with the equations in the previous chapter.

14
 4.1.3.1 Ruling span method
Since the spans are coupled the sag can be approximated with the use of a so called
ruling (or equivalent) span. A ruling span is a hypothetic span that captures the
behaviour of the entire line section. The ruling span is defined as:

S 13 + S 23 + ... + S n3 (4.7)
RS = S 1 + S 2 + ... + S n

where

RS Ruling span for line section containing n spans

S 1 - Span length of first span
S n Span length of n : th span

The sag in any one of the spans can then be calculated as

2
S
Di = DRS i (4.8)
S RS

The ruling span method can be used for reasonably level spans, see [3].

4.1.3.2 Numerical sag-tension calculations

A more accurate sag calculation of the sag can be done with numerical analysis. There
are several computer programs that can do this. In this project a program called Sagsec
was used. Sagsec lets one model the full line section and calculates, using finite
element analysis, the forces on each insulator and the resulting sags of the line section
can be determined.

4.2 Conductor temperature model

The model of heat gain in a conductor is derived from the general heat equation for a
homogeneous and isotropic solid [2]:

T 2T 1 T 1 2T 2T q(T , , z, r , t )
= + + + +
t c r 2 r r r 2 2 z 2 c (4.9)

15
 c = specific heat capacity (J/kgK)
q = power per unit volume (W/m3 )
r = radius (m)
T = temperature (K)
t = time (s)
z = axial length (m)
= mass density (kg/m 3 )
= thermal conductivity (W/mK)
= azimuthal angle (rad)

This model can be simplified with the assumption that the conductor has cylindrical
symmetry and semi-infinite length. This gives [2]:

T 2T 1 T q(T , , z, r , t )
= + + (4.10)
t c r 2 r r c

If the radial temperature distribution is neglected then Tav = Tcore = Tsurface.

Mass per unit length, m, and power per unit length, P, is

m = A
(4.11)
P = qA
where A is the cross-sectional area.

This reduces the conductor temperature model to a non-linear ordinary differential-

equation on the form [2]:

dT
mc =Pj + PS Pr Pw
dt (4.12)

where

Pj = current heating
PS = solar heating
Pr = radiative cooling
Pw = wind cooling

[2] also includes the magnetic heating, PM, in equation 4.12 but this term will be
considered to be included in Pj within this master thesis project.

Mass per unit length and heat capacity will be considered constant within this report.

16
The conductor heat storage constant mc is calculated from [2]:

mc = ma ca + ms c s (4.13)

where

ma = mass per unit length for aluminum section (kg/m)

ms = mass per unit length for steel section (kg/m)
ca = specific heat capacity for aluminum (J/kgK)
cs = specific heat capacity for steel (J/kgK)

The values of mass and heat capacity are taken from tabulated values [2],

The cooling and heating terms of the model will now be described individually.

4.2.1 Current heating

The size of current heating depends on the size of the load current and the resistance at
the prevailing conductor temperature. Magnetic heating and the heating due to skin
effects also affects the size of current heating. [2]

The DC resistance of a conductor depends on the conductor area, the conductivity of

aluminum, the lay lengths (the axial length of a whole turn in the spiral for the strands)
and temperature. [6].

The AC resistance is higher than the DC resistance due to the so called skin effects.
Skin effect refers to the tendency of the alternating current to flow near the surface of
the conductor, this reducing the effective area and hence increasing the resistance. The
skin effect increases with increasing frequency [11], decreases with increasing current
density and increases with conductor diameter [6].

The number of layers of aluminum in the conductor affects the current heating
characteristics. The strands are spiraled in opposite directions for every layer to avoid
unwinding of the wire. This means that odd layer conductors behave differently than
conductors with an even number of layers. [6]

An approximate method for calculating the current heating for steel cored conductors
is [2].

2
Pj = k j R AC (1 + (T 20)) I AC (4.14)

where

k j = estimated parameter
R AC = AC restistance at 20C ()
= temperature coefficient of resistance per degree (K -1 )
I AC = AC current (A)
17
In this project a tabulated value for the AC resistance at 20 degrees has been used and
the value for the constant kj been obtained through model estimation. This constant is
thus assumed to take care of skin effects and conversion from DC to AC current and
other effects. For a more complex current heating model, see [2].

4.2.2 Solar heating

The conductor is heated from solar radiation. The amount of heat gain from solar
radiation depends on the diameter of the conductor, its inclination to the horizontal
plane, the absorptivity of the conductor, the direct solar radiation, the diffuse sky
radiation and the reflected radiation. [2]

The measuring equipment for diffuse and direct solar radiation is however expensive
and needs regular attention which makes them difficult to use on remote sites.
Measuring equipment for global solar radiation is cheap and reliable and is therefore
more commonly used in these types of projects, including this. This means that the
model is simplified to only depend on absorptivity, the solar intensity and the
conductor.

PS = s SD (4.15)

where

s = absorptivity
S = solar radiation (W/m 2 )
D = conductor diameter (m)

The absorptivity is the fraction of the incoming solar radiation that is absorbed in the
conductor . This value varies within the lifetime of the conductor. As the conductor
gets older this value increases. There are different opinions in the literature about
which values of absorptivity that should be used. [2] states that an absorptivity of 0.5
can be used for most purposes, together with an emissivity of 0.5 (see also following
section 4.2.3).

4.2.3 Radiative cooling

All bodies emit energy in the form of electromagnetic radiation [14]. The emissivity of
a body is the fraction of the emitted energy and the energy a black body would have
emitted at the same temperature. The emissivity depends on the material of the

18
surface, temperature, surface condition and wavelength distribution of the emitted
energy. [1] The wavelength distribution for electrical conductors is mostly in the
infrared spectrum. The emissivity of conductors increases with the age of the
conductor. Empirical studies on ACSR conductors have shown that newly installed
conductors can have an emissivity of 0.23 while old conductors can have an emissivity
of 0.95. [1] The age dependence of the emissivity for ACSR conductors can be
described by

0 . 70 Y
=0 . 23 + (4.16)
1 . 22 + Y

where Y is the age of the conductor in years.

The relation is also shown in Figure 4.4.

Figure 4.4: Emissivity as a function of conductor age according to equation 4.16

The total radiative cooling can be approximated with [9]

Pr =
2
{ [ 4 4
] [
D g Tc Tg + d Tc4 Td4 ]} (4.17)

where

= Stefan - Boltzmann' s constant

g = effective emissivity for conductor facing the gruond
d = effective emissivity for conductor facing the sky
Tc = temperature of the conductor in Kelvin
Tg = temperature of the ground in Kelvin
Td = temperature of the sky in Kelvin

19
The emissivities are different for the top part of the conductor that faces the sky and
the lower part that faces the ground. However, since the radiation loss is a small part
of the total heat loss it is sufficiently accurate to use the overall ambient temperature,
Ta [9].

4 4
Pr = D (Tc Ta ) (4.18)

The emissivity and absorptivity of the conductor are closely related. Kirschoffs
radiation law states that the monochromatic absorptivity (s) and emissivity (s) of a
surface are equal, s = s. The incoming solar radiation lies mostly in the short
wavelength spectrum and the radiation from the conductor mostly in the infrared.
Therefore the infrared emissivity is used in (4.18). The short wavelength emissivity,
and therefore the absorptivity according to above, can be approximated by: [1]

s = s = + 0.2 (4.19)
s = s 1

4.2.4 Wind cooling

As the conductor due to current and solar heating usually is warmer than the
surrounding air the conductor heats up the air adjacent to its surface. If there is a wind
the heated air is carried away and new cold air can be heated by the conductor. Since
heat leaves the conductor this has a cooling effect on it. If there is no wind there is still
some measure of cooling since the density of the air heated by the conductor decreases
and the warm air slowly rises. New air takes its place and this has a slow cooling
effect on the conductor. In the case of no wind this is known as natural convective
cooling and with wind is known as forced convective cooling.

In [7], three formulas are suggested for calculating wind cooling, one for no wind
(qcn), one for low wind speed (qc1) and one for high wind speed (qc2). The air density,
dynamic viscosity of air and thermal conductivity of air depend on the film
temperature, which is the average between the conductor and the surrounding
temperature, and values for these for different film temperatures are tabulated in [7]
Of the three equations the one that gives the highest value at the present circumstances
should always be used. The formulas in [7] are:

D f Vw
0.52

qc1 = 1.01 + 0.0372 k f K angle (Tc Ta ) (4.20)

f

0.6
D f Vw
qc 2 = 0.0119 k f K angle (Tc Ta )

f

20
 (4.21)

0.5 (4.22)
q cn = 0.0205 f D 0.75 (Tc Ta )1.25

where

f = air density
f = dynamic viscosity of air
k f = thermal conductivity of air
K angle = wind direction factor (1 for perpendicular wind)
Vw = wind speed

In [2] two formulas are suggested, one in the case of no wind and one for wind. These
formulas are expressed using the dimensionless numbers Reynolds number, Grashof
number, Nusselt number and Prandtl number.

For this project the modeling of convective cooling was simplified to

Pw = k w (Tc Ta ) V n (4.23)

where

k w = estimated parameter
V =perpendicular wind speed (m/s)
n = estimated parameter

and the impact of natural convection was overlooked.

When measuring wind speed the standard is to measure ten-minute-mean values for
the wind speed. This means that a lot of small changes in wind speed and wind angle
are lost in the data. To avoid this loss of information a more accurate wind speed
measurement equipment was installed that collects data every second, which is then
used in the model as one-minute-mean values. The cooling effect from the wind is
assumed to be only from the perpendicular wind so the data is first recalculated to a
cooling wind speed which is the perpendicular resultant wind vector. This is the
value of the wind speed that is used throughout this master thesis report.

The anemometer measures the wind speed and direction ten times per second and
calculates a one-second-mean value for the vertical wind speed and the horizontal
wind speed and direction. Figure 4.5 shows the coordinate system used to calculate the

21
cooling wind speed. Wind components in the X direction are parallel to the line and
will not contribute to the cooling of the line.

Figure 4.5: Coordinate system for calculation of cooling wind speed. Only wind
components in Y and Z direction will contribute to cooling.

The horizontal wind speed vector is divided into perpendicular and parallel wind and a
one-second-mean value for the cooling wind speed is calculated as:

Vsec = VY2,sec + VZ2,sec (4.24)

where

Vsec - One - second - mean value of cooling wind speed

VY ,sec - One - second - mean value of wind components in Y - direction
VZ ,sec - One - second - mean value of wind components in Z - direction

A one-minute-mean value of Vsec is the cooling wind speed V used in this thesis.

4.3 Thermal rating

Overhead lines that are thermally limited are designed and installed to have enough
electrical clearance for all circumstances when the conductor temperature does not
exceed its temperature limit. This temperature limit can be different for different lines.
For these lines that are examined within this project the conductor temperature limit is
50C.

22
The thermal rating of an overhead line is the calculated maximum allowed current.
Thermal rating is calculated using:

Pr (Tmax ) + Pw (Tmax ) PS
I rating = (4.25)
R(Tmax )

This formula is derived from the conductors steady-state heat balance.

4.4 Line uprating

Line uprating is the process of making more capacity of a line available to use. Static
rating is the traditional method of rating overhead lines. Static rating uses conservative
assumptions about weather conditions. Common assumptions are wind speed of 0.6
m/s and full summer insolation. Assumptions about ambient temperature can vary and
some utilities use seasonal static ratings with different ambient temperatures for
different times of the year. At Vattenfall the ratings are based on an ambient
temperature of 30C in summer and 10C in the winter for 130kV lines.

As the electricity market was deregulated the uprating of overhead lines has become
more popular. There are several ways to uprate existing overhead lines. [5]. Some
uprating methods include physical changes of the line, such as building higher towers,
others are simpler, as the ones described below.

4.4.1 Redefining weather conditions

The static ratings are usually much lower than the maximum possible rating. The
easiest way of uprating is to just redefine the worst-case weather assumptions to less
conservative by assuming a higher wind speed. This is however a risky method as
field studies have shown that the wind is very low much more often than has been
previously believed. [5]

4.4.2 Dynamic Rating and real-time monitoring

Dynamic rating is a method that allows increased line load since it makes it possible to
utilize the full capacity of the lines instead of being limited by a worst case scenario-
approach. Dynamic rating requires measuring of several weather parameters at remote
sites and a communication system that allows the data to be sent to the operations
center. [5] The rating is calculated from actual weather conditions and is thus usually
higher than the static ratings.

23
Real-time monitoring of sag, tension or conductor temperature is easier than dynamic
rating based on real-time weather conditions. However, real-time monitoring demands
more of the system operator. As where dynamic rating provides the operator with an
allowed current load, simple real-time monitoring based rating requires the operator to
estimate how large the load can be based on the properties of that specific line. [5]

Dynamic rating calculations can be based solely on real-time weather data or include
calculations based on real-time conductor temperature, sag or tension data. If sag or
tension are measured a correlation between them and conductor temperature can be
calibrated.

4.4.3 Probabilistic methods

Weather data can be used to determine the probability that the maximum allowable
temperature would be reached. A reasonably low risk could then be allowed and thus
the rating could be increased. There are different ways to uprate lines in this way, with
more or less complex ways. When doing this historical weather data can be collected
from some nearby site. The methods can include the risk of an accident and the risk of
an object or person being under the line. The methods can also include the load profile
of the line and be combined with real-time monitoring. [5]

4.5 Correlation between weather parameters

The measured weather parameters within this project; solar intensity, ambient
temperature and wind speed, are not entirely independent. If the solar intensity
increases, the conductor temperature will increase, as well as the ambient temperature.
The insolation that heats the ground and air causes instability in the atmospheric
boundary layer which generates air movements (wind).

What this means in practice for the analysis is that it is often difficult to find periods
where only one parameter is changing.

For figures describing correlations between measured weather parameters see

Appendix 1.

4.6 Grey-box modelling

In empirical modelling of unknown systems one can take one of three different
approaches, so called white-box, black-box or grey-box modelling. To search for a
general physical model from known physical relationships is called white-box

24
modelling. To estimate a model based on measurement data without physical insight is
black-box modelling. In between these two extremes is grey-box modelling which is
searching for a model using the known physical parts of the system and using
measurement data to estimate the unknown parts. Grey-box modelling will be used
within this thesis for modelling of the line temperature. Grey-box modelling has the
advantage that what is already known of the system is utilized and that the estimated
parameters can have physical meaning. [18]

When building models it is natural to make approximations and simplifications. It is

not meaningful to build a model that is more exact than is required from the purpose
of the model. A simpler model with fewer variables gives shorter execution times for
simulations in a computer.

For dynamic models the time constants of the modelled system describe how fast the
system reacts to changes in the input variables. If the time constants in a system are
very different the differential equations are called stiff and require particular
numerical solvers.

When estimating the parameters in a dynamic model the main method is to minimize
the prediction errors. The minimization method used within this thesis is the Output-
Error method where the modelled output is compared to the measured value.

25
5 Methods
5.1 Data analysis

The analysis of the measurement data was at first approached as to find a completely
unknown model of the system. This first approach was to make a multiple linear
regression model of the line temperature. The second approach was to take the
physical model found in literature and estimate the parameters to the measurement
data (grey-box modelling).

The analysis of the data was done on data from line ZL8. For analysis of OL9, see
section 5.2.

5.1.1 The system

The system that was to be examined is illustrated in Figure 5.1. This representation of
the system shows that the parameters affecting line temperature are current, insolation,
ambient temperature and wind speed. The line temperature in its turn affect the line
sag. The dotted lines in the figure illustrates the correlations between insolation and
ambient temperature and wind speed.

Figure 5.1: Representation of the system describing internal relationships

between system components according to regression model

The system can also be described in functions of the cooling and heating as in Figure
5.2. In this figure the complexity of the system is visible where the only factor that

26
affects the line temperature independently of the line temperature itself is the solar
heating. This figure describes the system according to the dynamic model.

Figure 5.2: Representation of the system describing internal relationships

between system components according to dynamic model

5.1.2 Sag-temperature relationship analysis

The sag-temperature relationship was analysed with three different approaches. The
purpose was mainly to see whether the correlation was predictable and slow or if it
would be non-linear and for example react faster at higher temperatures.

The correlation between measured distance to ground and measured line temperature
was examined in plots for measurement periods July-October 2010 and April 2011. A
simple linear regression curve was calculated with the Matlab function regress.

To see how much the 40 kg measurement system hanging on the line affected the
behaviour and also to see how the coupled spans-effect affected measurements the line
section was modelled with the computer program Sagsec with and without weight of
measurement box and also for different temperatures. The measures of heights of
attachment points and span lengths of the line section were taken from tables and from
line-profiles. Since the total line section was long and finding properties of inclination
had to be done by hand with a ruler on the profile images only the spans closest to the
measurement span were modelled exactly and the rest of the section was approximated
using the ruling span concept. The results for vertical position of the modelled weight

27
(representing the measurement box) were then plotted versus temperature change and
compared to measurement data.

For calculations in Sagsec the values for tension and elasticity modulus used were the
assumed final values for tension and elasticity modulus after maximum ice and wind
load. The values used in the Sagsec-calculation were tension of 41.56MPa for 0C, for
an E-modulus of 60000MPa [13] and for an initial tension of 50MPa.

The approximate equation for sag as a function of span and line length (equation 4.5)
was used with the linear thermal elongation model and the resulting sag-temperature
relationship was compared to the Sagsec-calculated and the measurement-based
relationship. The distance to ground measurements were translated into sag using
proportions measured on the profile image of the span.

5.1.3 Modelling line temperature with a multiple linear regression model

The first attempt of a line temperature model was to make a multiple linear regression
of the line temperature dependence of ambient temperature, current, insolation and
wind speed, based on the relations in Figure 5.1. The regression was performed with
Matlabs function regress.

A problem that did arise when regression was performed in Matlab was the
multicollinearity within the data. The high condition number of the data matrix
indicates that the model is sensitive and not suitable to draw conclusions from about
the physical behaviour of the model components. One of the objectives with the model
was to be able to compare the magnitude of impact the different parameters had on the
line temperature. The multicollinearity therefore makes the regression model not very
useful, even though a good fit was achieved.

5.1.4 Calibrating the dynamic model of line temperature in Matlab

As was found in literature [2], [7] the system is better described as in Figure 5.2 with
dynamics. A grey-box model was therefore estimated using the structure of the non-
linear ordinary differential equation described previously (section 4.2) The unknown
parameters to be estimated were kj, kw, and n. The Output-Error method was used for
parameter estimation. This means that the output of the model, the line temperature, is
compared to the measured line temperature in search for the best model fit.

28
The unknown parameters to be estimated are:

k j
(5.1)
= k w
n

The model error (prediction error) is defined as:

e(t , ) = y (t ) y (t , ) (5.2)

where

y (t ) measured line temperature at time t

y (t , ) model output at time t

The unknown parameters are found by the following minimization:

= arg min V ( )

(5.3)
V ( ) = e 2 (t , )

where

- the minimizing
V the loss function

Matlab includes several built-in ordinary differential equation solvers. For this
purpose the solver ode15s was used, which uses a backward differentiation technique.
The s in the name stands for stiff which means that the solver changes the step
size depending on how fast the system is changing. A couple of other solvers were
also examined on the model but seemed to be less accurate (ode45, ode23s). The
possibility of using the simpler model as Euler forward was also examined but was
slower and less accurate than the built-in solvers.

29
The ode-solvers in Matlab require continuous input data so the sampled measured data
had to be interpolated before integration. The interpolation was done with the Matlab
function pchip.

The optimization of the model was performed by using the Matlab function
fminsearch which is a function that alters the specified coefficients and searches for a
minimum of some function. The function that fminsearch should minimize was set to
be the sum of square errors (equation 5.3). This means that in every iteration Matlab
solves the differential equation for a new set of coefficients and compares this to the
measured line temperature and decides if the square error is smaller than in the last
iteration.

Parameter estimation using a calculated derivative from the equation to the measured
line temperature derivative for every data point was also examined but failed as it was
difficult to get a good estimation of the measured derivative.

The problem with optimization of non-linear functions is that there can be several
local minima that cannot be separated from the global minima. This means that the
result from optimization is not necessarily the best one.

Several attempts were made to estimate all the coefficients (including s and ) but
without result as the Matlab program froze after too many iterations. The optimization
process therefore needed some manual testing of different combinations of
coefficients and getting a feeling of which changes led to a better model. This led to
an assumed higher absorptivity and emissivity, based on the age of the conductor (see
section 4.2.3). The assumed emissivity for the 50-60 years old line was 0.91 and the
corresponding absorptivity was estimated to be 0.99, since the absorptivity should be
higher than the emissivity though not bigger than one. Optimization using the
procedure described above could then be performed to find kj, kw and n.

The model was calibrated on measurement data from a couple of days in September
2010 and validated on data from April 2011.

5.1.5 Evaluation of dynamic line temperature model

Residuals were calculated and examined versus all the model parameters. For the
model to be good there should be no patterns in the residuals, they should for example
not be bigger for high currents than for low. This was used as a tool in the
optimization process and was one of the reasons that the impact of insolation seemed
to be underestimated at first with the low value of absorptivity.

30
5.1.6 Examination of the system behavior using the dynamic model

The dynamic model and the developed Matlab code were used to simulate different
scenarios of interest. Step responses of current, wind speed and insolation were
examined and time constants calculated from the results. The time constants were
calculated as the number of minutes after the step change before the line temperature
exceeds 63.2% of its final value. Extreme worst-case scenarios were modeled for the
two lines. The contribution to the temperature change of the different heating and
cooling terms was modeled using the coefficients derived in optimization of the
dynamic model and the actual weather data and visually compared to the measured
line temperature and the modeled line temperature. An estimation of the actual rating
for a period was performed using actual weather data and the maximum allowed
temperature assuming steady state for the rating.

5.2 Estimating a dynamic line temperature model for OL9

As the line temperature data for OL9 is missing a model could not be fitted for this
line. In order to still try and get something out of the existing data and to be able to
compare the two lines a model was estimated using the coefficients from the model for
ZL8.

The two lines differ in diameter, steel-to-aluminum-ratio and resistance. These known
differences were inserted in the model for OL9. The curve fitted coefficient in the
current heating term, kj, was assumed to be the same for OL9 as for ZL8. The solar
heating and radiative cooling term for OL9 were assumed to have the same
absorptivity and emissivity as ZL8. The simplified wind cooling term coefficient kw
for OL9 was assumed to be:

n
DOL 9
k w,OL9 = k w, ZL 8 n (5.4)
DZL8

with n = nZL8.

The line temperature that was modelled with these assumptions was then compared to
the distance to ground measurement. Since the direct relation between distance to
ground and line temperature not is known apart from a relative estimation no
validation could however be done. The results from the OL9 data modelling must
therefore be seen as very uncertain.

31
6 Results
6.1 Sag versus line temperature, ZL8

Figure 6.1 shows the measured values for distance to ground plotted versus the line
temperature for measuring period July to October 2010. See appendix 2 for measures
of the span.

Figure 6.1: Measured distance to ground versus line temperature for July to
October 2010, ZL8

The same relation for measuring period April 2011 is shown in Figure 6.2. The bigger
spread of the data in Figure 6.1 can at least partly be explained by measurement
difficulties. Figure 6.2 shows a linear relationship between distance to ground and line
temperature for temperatures between 0C and 35C.

Linear regression performed on the April-data results gives equation 6.1.

Z = 0.037 Tc + 13,11 (6.1)

where

Z = distance to ground (m)

Tc = conductor temperature (C),

32
with an R2-value of 0,97.

This means that the model implies a decreasing distance to ground with 3,7 cm per
degree Celsius.

Figure 6.2: Measured distance to ground versus line temperature for April 2011,
ZL8

7
Sagsec without box
6 Sagsec with box

5 Best linear fit from measured

Sag (m)

Sag from Slack at Midpoint

4

0
0 10 20 30 40 50 60
Temperature

Figure 6.3: Four curves describing sag versus line temperature, derived with
different methods.

The dark blue and the purple curves in Figure 6.3 shows the result of the Sagsec-
calculation for different temperature with and without impact of the weight of the
measurement box. The difference between the two curves indicates that the weight of

33
the box affects the measurements of ground distance with between 0,5-0,6 meters in
comparison to how the result would have been without the impact of the weight. The
both curves in Figure 6.3 show a linear relationsship within the temperature range of
10 to 50C.

Figure 6.3 also shows the result from sag-approximation for one isolated level-span
using equation 4.5 plotted versus different conductor temperature changes, assuming a
reference length of the conductor of 247,35 meters for 0 C. Slack is the difference
between line length and span length. All curves except the light blue describe the
lowest point of the catenary. The light blue curve describes the midpoint of the span
assuming a level span (see 4.1).

6.2 Result from multiple regression

Figure 6.4 shows the measured line temperature data and the regression model applied
on the weather data for period beginning 10th of April 2011 and 5 days ahead. The
calibration data period that was used was two days in September 2010. These results
were not used further, see section 5.1.3.

Figure 6.4: Regression model and measured line temperature for validation data
set

The model from regression is

34
Tc = 7.96 10 6 I 2 + 0.0106 S + 1.0041 Ta 0.0223 V (6.2)

where the line temperature Tc and ambient temperature Ta are measured in C and V is
perpendicular wind speed, I is current and S is insolation.

The condition number for the calibration data set is 6.23*105.

6.3 Calibration results for dynamic line temperature model

The final model was achieved with the values in Table 6.1. The estimated parameters
are kj, n and kw for ZL8.

Table 6.1: Values used in dynamic model for ZL8 and OL9
Line ZL8 OL9
D (mm) 39,2 27,8
Area (mm2) 910 454
2
Area S (mm ) 44 52
2
Area Al (mm ) 866 402
-1
alfa (K ) 0.00403 0.00403
Rac (/km) 0.081 0.037
kj 1.033 1.033
s 0.99 0.99
0.91 0.91
n 0.5373 0.5373
kw 2.066 1.7177

The result from validation on measurement period April 2011 of the dynamic model is
shown in Figure 6.5. Modelling for shorter periods are shown in Figure 6.6 (April 11
to April 14) and in figure 6.7 (first of April).

35
Figure 6.5: Modelled and measured line temperature for validation data

Figure 6.6: Modelled and measured line temperature for four days in April

36
Figure 6.7: Modelled and measured line temperature for one day in April

Figure 6.8: Distribution of model error for the April data

The model error distribution for the month of April is shown in Figure 6.8. Model
error is defined in section 5.1.4 and is the difference between measured line
temperature and model output.

The residuals of the model are plotted versus the measured parameters and displayed
in Figure 6.9. In Figure 6.10 the residuals versus measured line temperature are
plotted.

37
Figure 6.9: Model error versus measured parameters for data July-October

Figure 6.10: Model error versus line temperature for data July-October

38
Figure 6.11: Wind cooling effects versus perpendicular wind for model described
in [7] and for simplified model (equation 4.23)

The wind cooling term in the dynamic model is compared to the wind cooling terms
suggested in [7] in Figure 6.11. The conditions used for the comparison is a conductor
temperature of 45C and an ambient temperature of 25C. This gives a film
temperature of 35C and values for the dynamic viscosity of air, air density and
thermal conductivity of air are taken from [7] using the values for en altitude of 0
meter.

The green line shows the simplified wind term model used in this project. The IEEE
model suggests that the highest calculated wind cooling always should be used. That
means that the IEEE model suggests using the purple curve up to about 0.2 m/s, then
using the red curve until the blue curve crosses and then finally the blue curve.

39
6.4 The influence of weather parameters on conductor temperature
change

The sizes of the cooling and heating terms in the dynamic line temperature model are
shown in Figure 6.12 for measuring period April 11 to April 16 2011. The curves
represent the different terms, negative for cooling terms and positive for heating terms,
and are calculated from the dynamic model coefficients and from actual weather data
for these days. The actual measured line temperature has been used in the calculation
and not the modelled line temperature.

Figure 6.12: The size of cooling and heating terms for six days in April

The size of the wind terms variation with wind speed and line-ambient temperature
difference is shown in Figure 6.13.

40
Figure 6.13: Wind cooling effect versus wind speed for different temperature-
differences

6.5 Time constants and step responses

The line temperature step response for a current step from 200 to 800 amperes for
different wind speeds between 0 to 5 m/s simulated with the dynamic model is shown
in Figure 6.14. The time constants as well as the difference between end temperature
and start temperature increase with decreasing wind speed.

41
Figure 6.14: Line temperature response to step change in current for different
wind speeds (with ambient temperature 10C, insolation 300W/m2

Step response simulation to a step increase in insolation from 0 to 1000 W/m2 is

shown in Figure 6.15. The three curves in the figure represent different constant
current levels. The size of the temperature increase due to the insolation step is the
same for all three current levels as well as the time constants.

Figure 6.15: Line temperature response to step change in insolation for different
current levels (with ambient temperature 10C, wind speed 1m/s)

42
Line temperature step response to a step change in wind speed is shown in Figure
6.16. Two different step changes are simulated; one increasing wind speed from 0.2 to
5 m/s and one decreasing form 5 to 0.2 m/s. The temperature curves show that there is
a difference between time constants for an increasing step and a decreasing step
change.

Figure 6.16: Line temperature response to step changes in wind speed (ambient
temperature 10C, current 400A, insolation 300W/m2)

The time constants and the step changes that are displayed in Figures 6.14 to 6.16 are
also shown in Table 6.2. The table is colour-coded for emphasize on the variations and
the results. The purple cells in the table describes the value that is different for the
different curves, the green cells displays the resulting time constants and the blue cells
display the resulting line temperature difference between final and initial value.

43
Table 6.2: Step change simulations and results
Line Line
Amb Time temp temp Temp-
Wind temp Insolation constant start stop diff
Step in current (m/s) (C) (W/m^2) (min) (C) (C) (C)
200-800
Ampere 0 10 300 49 31 63 32
0,5 10 300 19 17 28 11
1 10 300 16 15 24 9
2 10 300 10 14 20 6
5 10 300 8 12 16 4

Line Line
Amb Time temp temp Temp-
Step in Wind temp Insolation constant start stop diff
insolation (m/s) (C) (W/m^2) (min) (C) (C) (C)
0-1000 W/m^2 1 10 200 11 10 21 11
1 10 400 11 12 22 10
1 10 800 11 11 28 11

Line Line
Amb Time temp temp Temp-
Step in wind Current temp Insolation constant start stop diff
speed (A) (C) (W/m^2) (min) (C) (C) (C)
5-0.2 m/s 400 10 300 24 13 22 9
0.2-5 m/s 400 10 300 6 22 13 -9

6.6 Worst-case simulation

Figures 6.17 and 6.18 show the differences in reaction time to a sudden current change
for the ZL8 line and the OL9 line respectively. The ambient conditions are set to give
a worst-case-scenario of constant wind speed at 0.2 m/s, constant insolation at
1000W/m2 and constant ambient temperature of 20C. The current step for ZL8 is
from 100 amperes to 700 amperes. This gives an initial line temperature of 45 C and
a final line temperature of 57 C. The temperature limit of 50 C is also shown in the
figure as the red line. The comparison between the two lines is made for different
current steps that result in the same initial and final line temperature. The temperatures
differ between the lines due to differences in resistance and diameter. The step change
for the thinner line OL9 that gives the same temperatures as the previously mentioned
step for line ZL8 is from 200 to 450 amperes. The temperature limit is reached in 13
minutes for ZL8 and in 8 minutes for OL9.

44
Figure 6.17: Time to reach temperature limit for ZL8 with constant worst-case
ambient conditions and sudden change in current

45
Figure 6.18: Time to reach temperature limit for OL9 with constant worst-case
ambient conditions and sudden change in current

6.7 Possible rating of the two lines

Figure 6.19: Possible and static rating for ZL8 during April

46
The possible rating calculated for every minute from equation 4.25 using measured
weather data is displayed in Figure 6.19 for ZL8 and in Figure 6.20 for OL9 for
available data in April 2011. The static summer rating for the two lines is the red line
in the figures. The green line shows the measured current during that time.

Figure 6.20: Possible and static rating for OL9 during April

47
7 Discussion
7.1 Analysis of ambient conditions and line temperature

A dynamic model has been estimated using the known physical properties of the
system as described in literature and coefficients estimated for the two lines.
Simulations using the dynamic model have been made to find time constants and line
temperature behaviour at worst case and the contributions to heating and cooling from
different weather parameters have been calculated based on the model.

The analysis within this thesis has not been to statistically look at all the available data
and see how often certain scenarios happen but rather instead to find the behaviour
and interactions between the input variables, illustrate and examine these behaviours
and create and simulate interesting scenarios. There could thus be a lot of information
remaining in the measurement data that could be analyzed in other ways.

7.2 The usefulness of the line temperature model

The developed dynamic model can be used further to see under which circumstances
the rating is the lowest and to simulate how often this would happen if new wind
power were connected to the grid resulting in higher currents on the line.

7.3 Possible weaknesses of the dynamic line temperature model

The dynamic model shows a good fit for the measured validation data set. The model
error distribution shown in Figure 6.8 shows that for the validation data set the error
was seldom bigger than 2.5 degrees. As seen in Figure 6.8 and also in Figure 6.9 there
are some outliers with a big model error. The cause of these deviations has not been
determined but a possible explanation is that the big errors occur at those times where
bad data have been excluded from the data set and the model needed time to adjust.

The most interesting scenarios within this project are the line temperature behaviour at
high currents and at resulting high temperatures and unfortunately there are not many
data points that fulfil those criteria. This means that the model could have a worse fit
for higher temperatures and/or higher currents than those within the validation data
set.

7.4 Implications from results of possible ratings

The possible rating shown in Figures 6.19 and 6.20 show that there was often a large
margin between summer static rating and possible rating for April 2011. This supports

48
the belief that there is a lot to gain by changing the rating method for these lines. This
analysis could be done for other periods as well for a fuller indication of the potential
rating. The most interesting periods to examine would probably be the warmest and
the coldest seasons.

Those few occasions when the possible rating gets close to the static rating should be
further analysed when deciding on how to proceed with different rating methods at
Vattenfall. If these occasions would show to appear due to low wind speed that also
could be correlated with low wind speed at nacelle height the low rating would not be
a problem for increased wind power connection.

The calculation of the possible rating is based on the assumptions of emissivity and
absorbtivity that was described in section 5.1.4. The emissivity of 0,91 coupled with a
high absorptivity of 0,99 gives a more conservative line temperature than a low
emissivity (0,5) that has been suggested (see section 4.2.2) for the measured line
temperatures as the insolation contribution to the line temperature is bigger than the
radiative cooling in everyday measured scenarios. However, when calculating the
rating of the line the maximum line temperature (50C for these lines) is used and this
means that the radiative cooling term gets very big as soon as the ambient temperature
is quite low and this effect gets bigger than the usually bigger insolation term that is
not temperature dependent.

7.5 The simplified wind model term

The simplified term for wind modelling compared to the complex wind terms as
suggested in other literature (see section 4.2.4) shows to come close to the low wind
term (Figure 6.11), for the examined temperature. This indicates that the simplified
model wind term, when isolated from the model as whole, describes a more
conservative wind cooling contribution than other models for higher wind speeds.

The figure also gives a sense of how much the natural convective cooling that is
overlooked in the dynamic model would have contributed to cooling. The figure
indicates that the theoretical natural convection would guarantee a perpendicular
wind speed of 0.2 m/s at all times. This has however not been tested on the model or
on the measurement data.

7.6 Time constants and times for reaching temperature limit

The examination of time constants and step responses had two important purposes.
One was to examine the impact of different line properties on the heat storing capacity
and thus the reaction time. The other was to estimate how fast an operator has to be

49
able to cut down the wind power production when the line temperature suddenly rises
due to sudden changes in order not to exceed the temperature limit.

The results from these simulations, as seen in Figures 6.17 and 6.18 implies that the
operator would have about 10 minutes to react to a sudden increase in current at a
worst-case ambient condition with a high initial temperature. For implementation the
control center would thus have to have a scheme that allowed them to act within 10
minutes after a sudden increase in current.

The results also indicate that a thinner line would result in faster reactions to current
changes. This would have to be considered if implementing any uprating method on
the lines.

7.7 Variations along the line

The measurements in this project have only been considering one point along each of
the lines. The points were selected with one criteria that they would represent a hot
spot, that is a point at the line likely to show the most extreme high line temperatures.
Good wind shielding was therefore the criteria. An overhead line is usually not
straight and the terrain can differ a lot along the line. As the wind direction is an
important factor when deciding the magnitude of wind cooling it is likely that the
situation can differ quite a lot between different parts of the line section. This should
be considered in further analysis. There is also an ongoing debate within the field of
dynamic rating whether line temperature measured in one point is suitable for real-
time monitoring or if it is too local and tension and or sag should be measured instead.
This matter will be discussed in a report from CIGR coming fall 2011.

7.8 Sag-line temperature correlation

Both methods for calculating the sag-temperature relationship, as well as measurement

data, show that within the temperature range from 0 to 50C the relationship is
approximately linear. This is a good result for the project as it indicates that measuring
the line temperature would be quite enough to roughly know the position of the line. It
would not be necessary to know the exact position, as the design included buffers, but
it is good that the sag does not suddenly increase faster with higher temperatures or
acts in any other unpredictable way.

As for the agreement between measured distance to ground and theoretically

calculated there is a good agreement between the rate with which the distance
decreases with temperature but a worse agreement for the absolute distance to ground.
As seen in Figure 6.3 the four curves have approximately the same slope but different
intersection points. This could possibly be a result from uncertainty in the measures

50
manually done in the profile image (Appendix 2). When looking at the profile image
one should be aware that the vertical scale is ten times bigger than the horizontal. As
the horizontal position of the measurement system is not known, but assumed to be at
the lowest point of the curve, and the ground under is very inclined a few meters error
in the assumption of horizontal position could result in some vertical error.

7.9 Suggestion of uprating method

The deregulation of the energy market around the world has been the driving force for
line uprating. Prior to the deregulation the lines were designed with very generous
clearance buffers and the conservative weather assumptions were widely accepted, but
since then as the competition on the market gets tougher and the difficulties in
transmission construction increases the energy utilities have come up with new
solutions such as implementing real-time monitoring systems. [17]

An even newer trend is to combine the overhead line uprating with the increasing
construction of wind power plants. Studies on this are made for example in Germany
[16] and in England [15]. The wind power plants often need to be located where the
existing network is designed for small loads [15]. Instead of reinforcements of the
overhead lines the idea with this type of dynamic uprating is that strong wind gives
high power production as well as adequate cooling of the overhead lines. As a safety
measure the wind power production can also be turned off in case the temperature
limit is reached.

For an implementation of an uprating method at Vattenfall with the purpose of

inserting new wind power on old lines the results and experiences from this master
thesis project would suggest a real-time monitoring system that measures the line
temperature and has the possibility to cut down wind power production in emergency
cases. This suggestion is based on a few findings. One is that it does not seem
necessary to measure the sag of the line, as measurements as well as theoretical
studies show that there is a predictable linear relationship between the line temperature
and the sag. This is also supported by literature [5] that states that one can choose
which one of line temperature, sag and tension that is to be measured for real-time
monitoring as they can be translated to each other. However, information from the
coming CIGR report should also be considered, see section 7.7. Real-time
monitoring is suggested rather than dynamic rating after the experience that
measurement systems often can fail to deliver data and thus the fewer parameters that
need to measured, the better. This suggestion does however rely on the assumption
that there is a clear correlation between low wind speed at line height at hot spots
along the line and at nacelle height at the wind power locations so that emergency cut-
downs of wind power production would be a very rare occasion. That means that
further research must be done before any final conclusions on implementation.

51
Research will continue at Vattenfall with a pilot project for a real-time monitoring
system beginning August 2011.

52
8 References
[1] Black, W.Z., Rehberg, R.L., (1985). Simplified model for steady state and real-
time ampacity of overhead conductors, IEEE transactions on power apparatus and
systems, Vol. Pas-104, No. 10

[2] CIGR, (2002). Thermal behaviour of overhead conductors

[3] CIGR, (2007). Sag-tension calculation methods for overhead lines

[4] Corren.se, (2011-05-24). www.corren.se

[5] EPRI, (2005). Increased power flow guidebook

[6] IEEE, (1985). AC Resistance of ACSR Magnetic and temperature effects, IEEE
transactions on power apparatus and systems, Vol. PAS-104, No.6

[7] IEEE, (2006). IEEE Standard for calculating the current-temperature of bare
overhead conductors

[8] Kungliga vetenskapsakademien, (2011-05-24). www.kva.se

[9] Morgan, V.T., (1982). The thermal rating of overhead-line conductors Part I. The
steady state thermal model, Electric power systems research, 5 119-139

[10] Pacific cabling solutions Ltd, (2011-05-24). www.pacificcabling.com

[11] Schavemaker, P., van der Sluis, L., (2009). Electrical power system essentials,
John Wiley & Sons Ltd, England

[12] Svensk energi, (2011-05-24). www.svenskenergi.se/sv/Om-el/Elnatet/

[13] Svenska elektriska kommissionen, (1999). Dimensionering av friledningar fr

starkstrm Ledare, Svensk standard SS 436 01 02

[14] Young, H., Freedman, R., (2004). University Physics, 11th edition, Pearson
Education, San Fransisco

[15] Yip, T., Chang, A., Lloyd, G., Martin, A., Ferris, B., (2008). Dynamic line rating
protection for wind farm connections, Developments in Power System azection, 693-
697

53
[16] Ringelband, T. et al (2009). Potential of improved wind integration by dynamic
thermal rating of overhead lines, PowerTech, 2009 IEEE Bucharest, 1-5

[17] CIGR (2006). Guide for selection of weather parameters for bare overhead
conductor ratings

[18] Glad, T., Ljung, L., (2004). Modellbygge och simulering, Studentlitteratur AB,
Lund

[19] Elskerhetsverket, (2008). ELSK-FS 2008:1, 7

54
Appendix 1
Measured line temperature plotted versus other measured
parameters

Figure A1.1: Measured line temperature versus measured insolation for period
July 2010 to October 2010

55
Figure A1.2: Measured line temperature versus measured perpendicular wind
for period July 2010 to October 2010

Figure A1.3: Measured line temperature versus measured ambient temperature

for period July 2010 to October 2010

56
Figure A1.4: Measured line temperature versus measured current for period
July 2010 to October 2010

57
Correlations between measured ambient conditions

Figure A1.5: Measured ambient temperature versus measured insolation for

period July 2010 to October 2010

58
Figure A1.6: Measured perpendicular wind versus measured insolation for
period July 2010 to October 2010

Figure A1.7: Measured perpendicular wind versus measured insolation for

period July 2010 to October 2010

59
Appendix 2
Profile image for measured span on ZL8

Figure A2.1: Profile image for measured span on ZL8. The span with
measurement system is in the middle.

60
Profile image for measured span on OL9

Figure A2.2: Profile image for measured span on OL9.

61